Meeting Agenda

Wednesday, May 7^tH, 1997

Location:

DC 1304

Time:

13:30

Chair:

Mark Riddell

1. Adoption of the Agenda - additions or deletions

2. Coffee Hour

Coffee hour this week:

Matthew Davidchuk

Coffee hour next week:

Volunteers?

3. Next meeting

Date:

May 14^th, 1997

Location:

DC 1304

Time:

13:30

Chair:

Liddy Olds

Technical presentation:

Dan Milgram

4. Forthcoming

Chairs:

1. Navid Sadikali (5/21)
2. Clara Tsang (5/28)
3. Teresa Yeung (6/4)

Tech Presenters:

1. Dan Milgram(5/21)
2. Liddy Olds (5/28)
3. Mark Riddell (6/4)

5. Technical Presentation

Presenter:

Marryat Ma

Title:

Using Perceptual Organization to Detect Corners In Scenes With Man-Made Objects

Abstract:

The detection and classification of three-dimensional (3D) objects within a scene using one or more two-dimensional (2D) images is extremely difficult. It is necessary to make a correspondence between the features in the 2D images and the object models in order to recognize the objects in the scene. Corner detection can be an extremely valuable tool in developing such a correspondence. Corners convey a great deal of information about objects within a scene. They are view-invariant features, which causes them to be especially useful for reconstructing 3D scene information from 2D images. In addition, corners serve to simplify the analysis of images by providing significant data reduction while preserving important information about the objects' structure.

In this presentation, I will discuss using perceptual organization techniques to detect corners in images containing man-made objects. The corner detection algorithm will be based on a curve detection algorithm that uses perceputal organization and is due to Qi-Gang Gao and A.K.C. Wong.

6.General Discussion Items

7. Action List

• Alias Brown Bag Seminar Series
• Ivan Sutherland Visit

8. Director's Meeting

9. Seminars

• Celebrity Lecture Series - Concurrency is Hard

Dr. Ivan E. Sutherland, Fellow and Vice-President, Sun Microsystems Labs
2:30-3:30 p.m.

DC 1302

Concurrent systems do many things at once. People find it hard to understand the rich patterns of activity that occur in complex concurrent systems. We find it even harder to tell in advance if a system can misbehave.

The clock in conventional computers simplifies the design task by grouping actions into clock periods. Given such a grouping, a designer need check only that all actions in each group are compatible, and that the precursors to each action occur in a prior group.

The clock, however, is a useful crutch over only a very narrow range of time scales. It offers no help for fine grain design where we must understand the analog behavior of each individual transistor. It offers no help for longer time scales such as found in operating systems where there are too many clock intervals to permit simple enumeration.

Why is concurrency hard? We seem to think sequentially, but is that inherent in the human mind or the result of decades of sequential programming. In circuit design and programming alike, much research today attempts to improve our ability to think about concurrency.

10. Lab Cleanup (until 14:30 or 5 minutes)